Reservoir characterization, machine learning and big data - An offshore California case study

Reservoir characterization, machine learning and big data - An offshore California case study

Big data

Offshore oil well production

Information management

Reservoir management

Supervised learning

Conservation

In order to robustly characterize a reservoir and make reservoir management decisions, it is paramount that an integrated and comprehensive study use all available static and dynamic data including petrophysical, geological, geophysical, engineering, and production data sets. These large vintage data sets are often available but are typically underutilized because of poor data management practices and lack of forward-looking data strategies. This paper presents the results of a supervised classification machine learning (ML) algorithm that accurately identifies reservoir quality associated with the most favorable production trends. The algorithm was trained and tested using log curves, seismic attributes, production, and sidewall core sample data sets. Lessons learned show the importance of managing data in a way that is complementary to machine learning. In addition, a flexible and forward-looking data strategy provides for rapid and efficient evaluation of reservoir characteristics. These quantitative machine learning results can be factored into field development strategies and help optimize efficiency and capital allocation. Integrating this machine learning workflow supports resource conservation efforts by ensuring optimal production of offshore hydrocarbon resources while minimizing impacts to the environment. Copyright 2020 Society of Petroleum Engineers.

Ojukwu, Chima

Smith, Kevin

Kadkhodayan, Nadia

Leung, Mark

Baldwin, Kimberly

SPE Nigeria Annual International Conference and Exhibition 2020, NAIC 2020, August 11, 2020 - August 13, 2020

2020

conferencePaper

10.2118/203642-MS

http://dx.doi.org/10.2118/203642-MS